[
https://issues.apache.org/jira/browse/SOLR-9555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15830078#comment-15830078
]
Mark Miller commented on SOLR-9555:
-----------------------------------
Comment was short because I was on my phone. Anyway, it's always a been a bug
the leader even tries to do this.
The way this should work is that replicas should set a watch on ZK LIR nodes so
they are notified when put into LIR - so that if they can speak to ZK but are
partitioned from the leader, they set themselves to not ACTIVE. In the
SolrCloud design, no replica or leader should ever publish state for another
replica or leader. Either the leader can talk to replica and puts it into
recovery, or it cannot and so communicates to it through ZK.
> Recovery can hang if a node is put into LIR as it is starting up
> ----------------------------------------------------------------
>
> Key: SOLR-9555
> URL: https://issues.apache.org/jira/browse/SOLR-9555
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Alan Woodward
>
> See
> https://jenkins.thetaphi.de/job/Lucene-Solr-master-Linux/17888/consoleFull
> for an example
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
